This thread is for community feedback and proposals around short term incentives for the FOAM Map.
Proposals for short term incentives can be put into two categories:
· short term intrinsic incentives
· short term extrinsic incentives
intrinsic incentives are hard coded into the smart contracts of the FOAM Protocol. (emphasis - these are not currently implemented) Currently we have not found a model that doesn’t also bring perverse incentives with it and that undermines long term token value. Some examples of perverse incentives caused by inflationary rewards are quoted at the end of this post.
extrinsic incentives are built on top of the core protocol as community resources or as part of the FOAM Map user interface. Examples include rewards, contests, bounties, leaderboards, reputation and mapathon events. To us, this category seems more promising for incentivising cartographers in the short term.
This thread is for discussion on both these types of proposals. For more reading on our view of incentives and the FOAM Map see this post. To kickstart the conversation, below follow some quotes from the post regarding perverse incentives.
Inflation punishes those Cartographers in these early stages that are taking their time to add quality points as well as physically visit points and document them before adding to the registry, in the case they need to present evidence in public forums as a defense in the case of a challenge. Inflationary rewards could perversely incentive quickly adding low quality points while diluting the shares of those being diligent.
If the inflationary reward pool for adding verified points is limited it could easily be drained by an attacker adding many low quality points quickly. This also applies if inflation was unlimited and rate limited per point.
If inflation for rewarding a verified point is time based there is always the n+1 problem of attack or accidental inaccuracies. For example if a point must be verified for 25 days to be eligible for a reward there is always the risk it is only found out to be invalid and challenged after receiving the reward. This also incentivizes hiding low quality points in obscure places.
If the inflation is too low, there is little added value. For example if for every 1 point added and verified you received 1 token, at the current minimum deposit of 50 tokens, you would be eligible for 1 “free” point for every 50. If the inflation is of any value of substance, it encourages perverse incentives of adding low quality points to chase rewards.
Most importantly, if the yield of staking on a point is of any significant value it exemplifies thoroughly an attack of a malicious actor accumulating tokens through inflation and defending their low quality points in challenges by voting heavily.
Quote from @Jacob on telegram: “Distribution of losing votes could be distributed locally, to all verified POI owners in a given radius. Those who build an area up are then rewarded. Hmmm, would that centralize a token holder’s dominance over an area?“
This idea seems nicely aligned with “ground truthing.”
I don’t know if it would centralise token holder’s dominance over an area. Maybe it would.
I suppose it could, potentially promote perverse incentive for inaccurate points if all the points around a maliciously inaccurate point are owned by the same malicious actor.
Or maybe it would create an “us” and “them” tribal mentality between global cities or disincentivize good curators from building outside their present radius or promote aggression toward anyone creating POIs from places they’ve been or lived before.
I like where you’re coming from though.
In the same way local businesses thrive when a good business opens near them. It would be nice if local POIs could benefit from the curation of strong POIs in their area.
Having said that, would it create elite POI zones and slum POI zones? That would be a worry.
As much as I like new ideas for different incentive dynamics, I keep coming back to agreement with the current, benign way the incentive mechanism around POIs work.
Especially now that I’ve read what can be built on top of the TCR to benefit good POI owners.
Thank you for thinking about this and putting your idea out there. It has spurred my exploration of “local incentive”. It seems like some element of local incentive could be quite cool on a smart contract layer. I will keep this in mind. Thank you.
The FOAM’s team rationale for not adding inflationary incentives is, for the most part, due to attack vectors and perverse incentives created by token rewards for POIs. I believe the FOAM team has spent a lot of time thinking about incentives, but I have not seen an adequate public discussion of this matter. Glad Arthur created this channel. Let’s discuss early-stage token reward incentive for cartographers here. I will strongly suggest everyone read Ryan’s article on incentives: “FOAM Map: Overview of the TCR Design and Incentives” (see Arthur’s link above)
I’ve been a dedicated follower of the FOAM project since I learned about it earlier this year. I have read both the technical and non-technical whitepapers, read all articles, attended meetups, and joined every community call since I’ve been in the community. I also tweet a lot about FOAM (cryptoZen) and subscribe to a majority of the US and international Telegram FOAM community channels. I participated in the token distribution event and currently have 24 POIs on the map. One recurring theme I’ve observed – before the token distribution and after – is the question of short-term incentives for cartographers. I believe short-term and early-stage incentives needs to be taken very seriously and prioritized given that FOAM is a “work-token” and success of the project is dependent on initial “work” by cartographers.
So, I want to start things off by describing a simple token reward distribution incentive for early-stage cartographers. Poke holes in it. If you don’t like it, say why. If you like it, say why as well. I’ve not 100% reflected on this but used some incentives and community-coordination wisdom garnered from cryptoland. I recognize the proposal below is very simple. The proposal by itself isn’t the whole point. The hope is that this discussion will be the basis of the community’s understanding of all factors at play with regard to cartographer’s reward (and perverse attacks) and also, this serves as a cradle from which other ideas may be born.
Note: All stated time-periods and percentages are for the sake of clarity and not written in stone. The FOAM team and community will decide exacts.
Early-Stage Reward Incentive (this will be a short-term intrinsic incentive) as defined by Arthur:
(1) Early-stage rewards are non-inflationary – that is, rewards be extracted from current token pool (see below).
(2) Like bitcoin, rewards should be on a linear (or low-power) decay function schedule over a defined time period – say 3-10 years. This is a bootstrap pool and expected to be fully drained in the ditto time period. The strategy here is simple. By end of the bootstrap period, the vision of a useful map on the blockchain would have been achieved and the incentivization to add POIs to the map will be different from todays.
(3) The token reward should be distributed equally among all existing verified POIs on the map with the caveats to be discussed below. This means that the early-stage cartographers (POIs) will receive more rewards than later cartographers.
(4) Rewards distribution:
(4a)Tokens rewards will only be distributed to verified POIs.
(4b) Snapshots of all verified POIs will be taken at pre-defined intervals i.e. bi-monthly or quarterly.
(4c) Verified POIs must have been on the map for a minimum of 8 weeks to be eligible for reward. This will mitigate the perverse act of adding low-quality or bogus POIs to the map as cartographers will have time to challenge such points. Also, since perverse attacks cannot be prevented, the reward distribution can only be designed to ensure “useful work” is performed by all POIs receiving rewards. With the 2-months eligibility period, even if an attacker’s POIs are only for the perverse purpose of receiving rewards, over the course of 2 months (or whatever the community decides), this attacker will have performed “useful work.” Also, the attacker’s POIs cannot be bogus or low-quality because they will be challenged given the long eligibility period.
(4d) No challenged points will receive reward distribution. To prevent a bogus-challenge-attack, POIs that win challenges and meet all other terms of distribution will receive their allotted rewards retroactively.
(4e) 25% (25,000,000) of the 10% of FOAM tokens “allocated to a Reserve intended for grants to developers, or other purposes related to use of the FOAM protocol” can be used for the early-stage rewards incentive.
(5) In order to make the token reward distribution as fair and un-gameable as possible, the following will have to be implemented into the FOAM map app:
(5a) Robust search feature (soon to be released)
(5b) Robust and flexible filtering features. At minimum, cartographers should be able to
(5bi) filter all POIs verified in the last 24 hours, 2 weeks, month, etc.
(5bii) filter all POIs that have never been challenged
(5biii) filter all POIs from new or never used wallets
(5biv) filter all POIs from a certain wallet address. This way, all POIs from a suspicious wallet can be easily filtered out and scrutinized.
(5c) Robust email/browser notification system.
(5d) A reputation system is put in place (I acknowledge this will be difficult to implement)
What exactly don’t you agree with? Be specific. This post is for those ready to think and be specific. To create uses for POIs, you need POIs. Right now, the short-term incentives to create POIs post PoU are low. Short-term incentives will catalyze the growth of POIs - making the map useful for whatever use cases come-up.
I really like this approach, I think it may be up to the project team to model out the exact numbers but its a very sensible way of doing it. The team have expressed concerns that incentives will lead people to add low quality POIs. I think this is a better problem to have then no engagement. At the current rate I believe a majority of users will make 10 POIs and stop engaging with the protocol. At least if you have low quality POIs you have a network of highly engaged participants challenging those points and taking the initial amount staked which should be far greater than the incentives anyway. Over time this would lead to highly engaged participants accumulating tokens off low quality contributors both through their contributions and challenges.
With the current incentive structure, I believe many rational actors will not want to add POIs because they are effectively burning 50 of their tokens for no financial gain, at least in the short term. Sure it may increase the value of the network over the long term, but rational actors will just free ride off the work of others.
Brilliant take on community engagement Nelson. Bogus/low-quality POIs will not necessarily be a bad thing. This gives the community a chance earn more tokens from challenges. Even better than earning tokens is an engaged and incentivized community. With the suggested long eligibility period and robust filters, search features, and reputation system, the effects of attacks will be minimized.
Trenton, I will have to think about this some more before a response. Reputation sounds great until one starts to get into the nuances of what it actually means and how to implement it so it fairly and effectively serves it’s purpose. Its a hard nut to crack. Even Ryan (Twobitidiot) of Messari admitted this much in his recent interview with Epicenter.
In the context of FOAM, given the global nature of POIs, reputation will serve as a quality filter - one of many. POIs submitted by highly reputable cartographers will be less suspicious and scrutinized than those submitted by new wallets or cartographer with low reputation.
Reputation can also be gamified in the context of NFTs. Think NFTs that can only be attained by cartographer who’ve attained certain reputation.
Another usefulness of reputation is reliability of POIs. Game developers and other users of POIs want to know POIs called by their contracts will remain available and durable. They don’t want their contracts calling POIs that will be deleted a week or 2 later later. Reputation can also be created around age and reliability of POIs submitted by cartographers. I have more but my day calls…
Thank you @cryptozen for your thoughtful points and takin the time to start the conversation around how to maximize the initial “work” done by cartographers.
Setting aside a pool of tokens specifically as a reward pool for short term awards is possible and addressed in my post. In the simple manner, it could be drained at once. As you describe, an optimal implementation would require the distribution to be time based and rate limited. However, there is still an incentive to add a low quality point to chase this reward. I believe this incentive is not fully mitigated in your model.
The scenario I have in mind that I am concerned about is that of geographic obscurity and disguised points that are explicitly difficult for token holders to verify with confidence. Lets say for example a malicious actor adds points to a region that no other token holder has reasonably visited, i.e. a small town in Mongolia. The malicious actor could add genuine points here, but also ones that are completely made up but appear to be reasonable. This attack works even better if there is no Google Street view to cross check. It is reasonable to believe that genuine points will be added to the FOAM Map that are not on Google Maps or OSM, but it is also reasonable to believe that no amount of windowed time for Cartographers to verify before rewards are distributed will be enough. A malicious point can always make it past the window of eligibility and receive this pay out, effectively “stealing” a portion from genuine points.
As I wrote in the blog post, even if the point is found out after having received a reward and is challenged it can be assumed any prepared malicious point creator would be prepared to defend challenges and their yield.
When there is no reward payout for points, having a malicious point in an obscure location is less of a problem. The point would likely be found out once there was sufficient activity in that region on the application level and be challenged. With reward payouts for human subjective work, I can not conceive of a model where the risk is completely eliminated for a point to receive a reward by gaming the system.
There are other means of increasing engagement without explicitly opening up the token distribution to rampant gaming. My stance is that rewards for subjective work incentives malicious actors. There is less of an incentive for malicious actors in the current model (while some still exist). Pursuing alternative means of increasing engagement is much more favorable than a system that could undermine the veracity of the map.
All of these are great and are planned to be added but are not mutually exclusive to a reward pool
Excellent points! Leaderboards and user statistics such as “is this challenger a large token holder?” “Does this user challenge points regularly?” Will serve as a form of reputation. Additionally NFT’s as badges for different achievements would be a great addition.
To clarify, adding a point is not equivalent to burning tokens. As a point owner, you retain ownership of the tokens deposited and can withdraw up to the minimum while keeping the point.
A major an important aspect of the FOAM protocol is governance. To start there is governance over the TCR parameters. This includes the minimum deposit, which can be lowered if needed and you would be able to withdraw additional tokens.
However, governance is not only about the parameters of the TCR but the future of the protocol at large, including a TCR 2.0 and implementing a smart contract upgrade. FOAM is a work token, but also a governance token that allows ownership and control over this consensus drive map. For example, here is some recent thinking from Chris Burniske:
I essentially agree in that building applications and tools that require POIs will drive demand and bring new points to the map and bring value. The more the FOAM TCR needs to be relied upon the more valuable it will become. There is a difference between virality and network effects. In the former, the goal is to get as many points added as quickly as possible, in the later value is robustly generated through the protocol having more users and more applications hooking into it.
As a web3 protocol and base infrastructure, the FOAM Map will hook into protocols such as 0x, Augur, Set protocol, Filecoin etc. It is open for any developer to build and capitalize on interoperability of an open financial system.
This is aligned with my current thinking as well. I will conclude this post with that in addition to the value of governance there is the intention of inflationary rewards to the computationally verifiable work of Signalers and Miners in Dynamic Proof of Location.
The model I am most attracted to at the moment is having the losing side of the vote or a portion of the reward pool go to all verified points, however this is quite infeasible with the current state of things due to the massive gas overhead this would entail. As a note, we do believe scalability solutions are coming and that gas fees will not be a blocker for the long term.
Thanks for the detailed response Ryan. To be clear, I’m doing this because you’ve challenged others to bring you viable cartographers reward models in previous Telegram conversations. I have always been a fan and will continue to be a fan of FOAM.
Argument for mitigation will be provided in subsequent paragraphs below.
There are always going to be attack vectors. However, a well-thought-out mechanism will render attacks benign or eliminate by making attacks unprofitable. I can list a few attack vectors for the TCR map as currently implemented today. It’s not that attacks don’t exist, it’s that they’ve been trustlessly pacified.
From your response, I take it that the only attack vector you’re concerned about with regards to my proposal is the “geographic obscurity” attack. Are there any others? Please feel free to list and describe them here.
I think the “geographic obscurity” attack as you described is a non-problem if one looks at things from this from the following lens: The FOAM map is a TCR – an objective TCR to be precise. What do we know about the characteristics of objective TCRs? According to Aleksandr Bulkin in this widely read and applauded article (https://blog.coinfund.io/curate-this-token-curated-registries-that-dont-work-d76370b77150 ), the following must hold for objective TCRs: “(1) the objective answer exists, (2) it is publicly observable, and (3) it is very cheap to observe it.” The “geographic obscurity” attack you described above does not meet (3) and it can be argued, within certain boundaries, does not meet (2) as well. So, geographically obscure POIs must be challenged on the grounds of “unverifiability.” It doesn’t matter if the POI is located in Mongolia, the Himalayas, or in the deep forest of West Virginia or Nebraska. If cartographers can’t verify the authenticity and correctness of a POI, then it must be challenged. This is not to say cartographers shouldn’t add geographically obscure POIs; it’s saying any POI added must be cheaply and publicly observable. If a POI cannot be found in Google Maps or OSM, then the cartographer will need to provide satisfactory proof of the POI’s authenticity and their familiarity with it. A good example is what Lief Eric Malone is doing with his POIs (https://twitter.com/eroxors/status/1046782615150374912 ). The FOAM team can formalize this. Again, from this point of view, the “geographic obscurity” attack is a non-problem.
I don’t agree “reasonable to believe that no amount of windowed time for Cartographers to verify before rewards are distributed will be enough.” My proposal has the following properties:
2 to 3-month POI eligibility period.
Robust search capability
Robust filter capabilities
Time-base and rate-limited distribution scheme
Time between reward distributions must be less than POI eligibility period.
The FOAM has said most of these features will be added.
Your position does not consider everything. The profitability of an attack in my model is bounded by the yield or amount of token gained from each token distribution event. An attacker is driven by simple math. If an attack is profitable, attack. If not, well, don’t attack. Well-reasoned distribution parameters (total tokens to be distributed per event, frequency of distribution, age of eligible POI, duration of bootstrap period, distribution decay function) will render wholesale attacks as you describe unprofitable.
Not sure why you use the word “rampant.” Do you have a model that suggest this?
Does this mean this mean the community can vote on early-stage token reward distribution?
What is meant by increased average half-life? You mean there is an increase in the length of time a point remains? Where does the “Half Life” come from?
That is quite a strong stance, is it not the same as guilty until proven innocent? In the scenario I laid out there is a mix of real and fake points, all equally unverifiable. All should then be challenged? What if the owner does not come forward with “evidence” - but they are still real points. You argue that unverifiable points should be challenged wholesale? I am not arguing for or against this, just clarifying. If that is your stance then geographic obscurity can be considered a “non-problem” but also seems to have baked in preconceived notions of how curators should engage with a TCR.
Not sure I follow the logic here, if there is a profit from distributed per event, frequency of distribution, age of eligible POI, duration of bootstrap period, distribution decay function for an honest point, why is it automatically unprofitable for an invalid point? The first question is “is the distribution scheme profitable period?”. Assuming yes, then there is incentive to search for an attack.
Geographic obscurity is not the only attack vector, but a subtle one highlighted in an example. I said rampant because I believe it will mobilizes “(b) low-quality points (although my proposal minimizes this, it might be a feature).” My point is that a distribution scheme opens up perverse incentives. Perhaps if curation efforts are ramped up because of this, instead of false points there are then a plethora of low quality.
I am not at all dismissing all together your proposal and agree that an added early stage distribution pool can catalyze points as well as token scarcity. However, that is not mutually exclusive to that scarcity catalyst coming from the demand for the points and registry from applications hooking into it.
With applications and adoption of the protocol these are also achievable. Governance rights will extend beyond the TCR, but so expectations are clear we are working on this 1.0 release and then the external incentive models described in the OP of this thread prior to a TCR 2.0 proposal.
Yes, I mean there will be an increase in the length of time a POI remains - especially during the early stages. Rewards for verified POIs will incentivize this.
Its a crypto coordination project. The #1 meme is “trust but verify.” If a POI is unverifiable, how can peers trust?
Yes, all unverifiable POIs should be challenged on ditto grounds
I will have to expand on this later. My day calls. But quickly, lets use an example. If a limitation is put on rewards received per distribution event i.e. rewards received cannot be more than 5% of minimum POI amount (with current parameters, this will be 5% of 50 FOAM = 2.5 FOAM per POI). The attacker will have to evaluate the cost/benefit ratio of the attack. Will the amount work and time required for an attack be worth the 5% yield?
Can you list other attack vectors for my proposal?
Yes, relative to ‘no incentives’, a distribution scheme opens up perverse incentives. However perverse incentives should not be considered in isolation. What advantages does a distribution scheme bring with it? and how effectively can the perverse incentives be stymied?
Low-quality is for the community of cartographers to decide on and challenge. In a twisted way, this can be a feature because cartographers can earn tokens from challenges. People love playing games. As it is now, what % of points are being challenged?
Noted. As I said initially, I had not seen an adequate discussion of early-stage token rewards. That’s what I set out to do. Now we have a public proposal. The viability of any new proposal can be compared to this. Also, the “rewards vs non-rewards” stance can be expanded on by the community.
Curious as I’m thinking through this: would a limited test of this proposal (or some similar variant) be feasible? I’m trying to imagine scenarios to explore this proposal while limiting the downsides, and it seems like a test limited by geography, time, and/or reward amount would help us understand some of the risks here. Or is there some reason we should we be thinking about these types of proposals as an ‘all or none’ sort of thing?
I had suggested in a Telegram discussion that the distribution of the losing side foam goes to the holders of validated POIs within a specific radius of the challenge. This would incentivize groups of people contributing to a wide range of areas because they would then pick up some foam from local challenges from several places.
This would effectively raise the quality of POIs as well in any given region. Suppose there is a region with numerous inaccurate POIs. The people who create them refuse to withdraw them. So they get challenged. Some FOAM from the loss actually goes back to those people in that region. Better POI contributors move in. Challenges are made, foam losses are increasingly distributed to new actors in the region.